diff --git a/docs/ISSUES.md b/docs/ISSUES.md
index 4655bdfc..9e33e213 100644
--- a/docs/ISSUES.md
+++ b/docs/ISSUES.md
@@ -24,6 +24,77 @@ What does NOT go here:
 - Every session: scan OPEN issues at start; promote/close anything we touched during the session before ending.
 - Promoting to a Phase: mark as `DONE (promoted to Phase X)` + commit SHA where the Phase entry landed.
 
+## #333 — The shadow broadphase reach filter measures from the PART ORIGIN, so an off-centre BSP part can be in the right cell and still never be tested
+
+**Status:** OPEN
+**Severity:** high for the tall-prop population. It is the gate immediately
+downstream of the AP-156 membership fix, so that fix alone may not be enough to
+make the worst objects block.
+**Filed:** 2026-08-06 at the AP-156 fix (commit `b52967de`), which surfaced it.
+**Do NOT bundle with AP-156.** Different code path (collision query, not cell
+membership), and it needs its own retail question answered first.
+
+### The mechanism
+
+`src/AcDream.Core/Physics/TransitionTypes.cs:3756-3764`, the shadow broadphase:
+
+```csharp
+Vector3 deltaToCurr = currPos - obj.Position;
+...
+float maxReach = sphereRadius + obj.Radius + movement.Length() + 2f;
+if (distToCurr > maxReach)
+    continue;                       // candidate discarded, never tested
+```
+
+`obj.Position` is the shadow row's PART placement —
+`entityWorldPos + rotate(shape.LocalPosition, entityWorldRot)` — while
+`obj.Radius` is the physics-BSP ROOT BOUNDING SPHERE radius, which is measured
+about that sphere's own centre, not about the part origin. AP-156 established
+that the two are frequently far apart: 376 of 973 installed physics-BSP parts
+have a root-sphere origin further from the part origin than half their own
+radius, worst 20.762 m on a 27.708 m sphere.
+
+Write `d` for the distance from part origin to the true sphere centre, `R` for
+`obj.Radius`, `r` for the mover's sphere radius. A mover just touching the
+geometry is `R + r` from the sphere's TRUE centre, hence up to `d + R + r` from
+`obj.Position`. The filter admits it only when
+`d + R + r <= r + R + movement + 2`, i.e. only when `d <= movement + 2`. Per
+physics tick the movement term is well under a metre, so any part whose root
+sphere sits more than about 2 m from its part origin can have a genuine contact
+discarded before `BSPQuery` ever runs.
+
+Worked case — Setup `0x02000255`, one part, root sphere
+origin `(0.000, -0.007, 9.911)`, radius `10.522`, `Setup.Height = 18.692`.
+`d = 9.911` against a budget of roughly 2.5 m. A player standing against the
+upper half of that prop is about 20.4 m from the part origin while `maxReach`
+is about 13.5 m. Discarded.
+
+### Why it matters now
+
+Before AP-156 those objects were usually not registered in the cell at all, so
+this filter never got the chance to reject them. AP-156 puts them into the
+correct cells; this filter is the next gate they hit. If the connected session
+finds a tall prop that still does not block after AP-156, look here first.
+
+### What to establish before fixing
+
+1. Does retail have this pre-filter at all? `CPhysicsObj::FindObjCollisions`
+   @0x0050f050 walks the cell's object list and dispatches per object; the
+   `+ 2f` slack and the `movement.Length()` term look like acdream's own
+   broadphase rather than a port. If it is acdream's, it also needs a
+   divergence-register row, which it does not currently have.
+2. If it is kept, it must measure from where the geometry is: `ShadowEntry`
+   needs the `BoundsCenter` that `ShadowShape` now carries, and `deltaToCurr`
+   must be taken against `obj.Position + rotate(obj.BoundsCenter, obj.Rotation)`.
+3. Cylinder rows are unaffected (`BoundsCenter == Zero` by construction) — the
+   change must not move them.
+
+### Test to write first
+
+A mover adjacent to an off-centre BSP part's geometry, with the part origin far
+outside `maxReach`. It must reach `BSPQuery`. Sabotage by restoring the
+part-origin measurement; it must go back to reporting no collision.
+
 ## #332 — Headless bots appear to have no remote dead-reckoning at all
 
 **Status:** OPEN (observation, not yet established as a defect)